aneka cloud application platform and its integration with windows
TRANSCRIPT
1
Chapter 27
Aneka Cloud Application Platform and Its
Integration with Windows Azure
Yi Wei1, Karthik Sukumar
1, Christian Vecchiola
2,
Dileban Karunamoorthy2 and Rajkumar Buyya
1, 2
1Manjrasoft Pty. Ltd., Melbourne, Victoria, Australia
2Cloud Computing and Distributed Systems (CLOUDS) Laboratory,
Department of Computer Science and Software Engineering,
The University of Melbourne, Australia
Abstract
Aneka is an Application Platform-as-a-Service (Aneka PaaS) for Cloud Computing. It
acts as a framework for building customized applications and deploying them on
either public or private Clouds. One of the key features of Aneka is its support for
provisioning resources on different public Cloud providers such as Amazon EC2,
Windows Azure and GoGrid. In this chapter, we will present Aneka platform and its
integration with one of the public Cloud infrastructures, Windows Azure, which
enables the usage of Windows Azure Compute Service as a resource provider of
Aneka PaaS. The integration of the two platforms will allow users to leverage the
power of Windows Azure Platform for Aneka Cloud Computing, employing a large
number of compute instances to run their applications in parallel. Furthermore,
customers of the Windows Azure platform can benefit from the integration with
Aneka PaaS by embracing the advanced features of Aneka in terms of multiple
programming models, scheduling and management services, application execution
services, accounting and pricing services and dynamic provisioning services. Finally,
in addition to the Windows Azure Platform we will illustrate in this chapter the
integration of Aneka PaaS with other public Cloud platforms such as Amazon EC2
and GoGrid, and virtual machine management platforms such as Xen Server. The new
support of provisioning resources on Windows Azure once again proves the
adaptability, extensibility and flexibility of Aneka.
Keyword
Cloud Computing, Platform–as-a-Service (PaaS), Aneka, Windows Azure, Dynamic
Provisioning, and Cloud Application Development
1. INTRODUCTION
Current industries have seen Clouds [2, 14] as an economic incentive for expanding
their IT infrastructure with less total cost of ownership (TCO) and higher return of
investment (ROI). By supporting virtualization and dynamic provisioning of resources
on demand, Cloud computing paradigm allows any business, from small and medium
enterprise (SMEs) to large organizations, to more wisely and securely plan their IT
expenditures. They will be able to respond rapidly to variations in the market demand
2
for their Cloud services. IT cost savings are realized by means of the provision of IT
"subscription-oriented" infrastructure and services on a pay-as-you-go-basis. There is
no more need to invest in redundant and highly fault tolerant hardware or expensive
software systems, which will lose their value before they will be paid off by the
generated revenue. Cloud computing now allows paying for what the business need at
the present time and to release it when these resources are no longer needed. The
practice of renting IT infrastructures and services has become so appealing that it is
not only leveraged to integrate additional resources and elastically scale existing
software systems into hybrid Clouds, but also to redesign the existing IT infrastructure
in order to optimize the usage of the internal IT, thus leading to the birth of private
Clouds. To effectively and efficiently harness Cloud computing, service providers and
application developers need to deal with several challenges, which include:
application programming models, resource management and monitoring, cost-aware
provisioning, application scheduling, and energy efficient resource utilization. The
Aneka Cloud Application platform, together with other virtualization and Cloud
computing technologies aims to address these challenges and to simplify the design
and deployment of Cloud Computing systems.
Aneka is a .NET-based application development Platform-as–a-Service (PaaS), which
offers a runtime environment and a set of APIs that enable developers to build
customized applications by using multiple programming models such as Task
Programming, Thread Programming and MapReduce Programming, which can
leverage the compute resources on either public or private Clouds [1]. Moreover,
Aneka provides a number of services that allow users to control, auto-scale, reserve,
monitor and bill users for the resources used by their applications. One of key
characteristics of Aneka PaaS is to support provisioning of resources on public Clouds
such as Windows Azure, Amazon EC2, and GoGrid, while also harnessing private
Cloud resources ranging from desktops and clusters, to virtual datacentres when
needed to boost the performance of applications, as shown in Figure 1. Aneka has
successfully been used in several industry segments and application scenarios to meet
their rapidly growing computing demands.
In this chapter, we will introduce Aneka Cloud Application Platform (Aneka PaaS)
and describe its integration with public Cloud platforms particularly focusing on the
Windows Azure Platform. We will show in detail, how an adaptable, extensible and
flexible Cloud platform can help enhance the performance and efficiency of
applications by harnessing resources from private, public or hybrid Clouds with
minimal programming effort. The Windows Azure Platform is a Cloud Services
Platform offered by Microsoft [5]. Our goal is to integrate the Aneka PaaS with
Windows Azure Platform, so that Aneka PaaS can leverage the computing resources
offered by Windows Azure Platform. The integration supports two types of
deployments. In the first case, our objective is to deploy Aneka Worker Containers as
instances of Windows Azure Worker Role, while the Aneka Master Container runs
locally on-premises, enabling users of Aneka PaaS to use the computing resources
offered by Windows Azure Platform for application execution. And in the second case,
the entire Aneka Cloud is deployed on Windows Azure so that Aneka users do not
have to build or provision any computing resources to run Aneka PaaS. This chapter
reports the design and implementation of the deployment of Aneka containers on
Windows Azure Worker Role and the integration of two platforms.
3
Figure 1: Aneka Cloud Application Development Platform.
The remainder of the chapter is structured as follows: in section 2, we present the
architecture of Aneka PaaS, provide an overview of the Windows Azure Platform and
Windows Azure Service Architecture, and list the advantages of integrating the two
platforms along with the limitations and challenges we faced. Section 3 demonstrates
our design in detail on how to integrate the Aneka PaaS with Windows Azure
Platform. Next, we will discuss the implementation of the design in Section 4. Section
5 presents the experimental results of executing applications on the two integrated
environments. In Section 6 and 7, we list related work and sample applications of
Aneka. Finally, we present the conclusions and future directions.
2. BACKGROUND
In this section, we present the architecture of Aneka PaaS, and then depict the overall
view on Windows Azure Platform and Windows Azure Service Architecture. We also
discuss the advantages brought by the integration, along with the limitations and
challenges faced.
2.1 Overview of Aneka Cloud Application Development Platform
Figure 2 shows the basic architecture of Aneka. The system includes four key
components, including Aneka Master, Aneka Worker, Aneka Management Console,
and Aneka Client Libraries [1].
The Aneka Master and Aneka Worker are both Aneka Containers which represents
the basic deployment unit of Aneka based Clouds. Aneka Containers host different
4
kinds of services depending on their role. For instance, in addition to mandatory
services, the Master runs the Scheduling, Accounting, Reporting, Reservation,
Provisioning, and Storage services, while the Workers run execution services. For
scalability reasons, some of these services can be hosted on separate Containers with
different roles. For example, it is ideal to deploy a Storage Container for hosting the
Storage service, which is responsible for managing the storage and transfer of files
within the Aneka Cloud. The Master Container is responsible for managing the entire
Aneka Cloud, coordinating the execution of applications by dispatching the collection
of work units to the compute nodes, whilst the Worker Container is in charge of
executing the work units, monitoring the execution, and collecting and forwarding the
results.
Figure 2: Basic Architecture of Aneka.
The Management Studio and client libraries help in managing the Aneka Cloud and
developing applications that utilize resources on Aneka Cloud. The Management
Studio is an administrative console that is used to configure Aneka Clouds; install,
start or stop Containers; setup user accounts and permissions for accessing Cloud
resources; and access monitoring and billing information. The Aneka client libraries,
are Application Programming Interfaces (APIs) used to develop applications which
can be executed on the Aneka Cloud. Three different kinds of Cloud programming
models are available for the Aneka PaaS to cover different application scenarios::
Task Programming, Thread Programming and MapReduce Programming These
models represent common abstractions in distributed and parallel computing and
provide developers with familiar abstractions to design and implement applications.
Client Libraries
Programming Models
1. Task Model2. Thread Model3. MapReduce Model
Master Container
Services
1. <core>2. Scheduling3. Accounting4. Reporting5. Reservation6. Provisioning7. Storage
Worker Container
Services
1. <core>2. Allocation3. Execution
Worker Container
Services
1. <core>2. Allocation3. Execution
Worker Container
Services
1. <core>2. Allocation3. Execution
Management Studio
5
2.1.1 Fast and Simple: Task Programming Model
Task Programming Model provides developers with the ability of expressing
applications as a collection of independent tasks. Each task can perform different
operations, or the same operation on different data, and can be executed in any order
by the runtime environment. This is a scenario in which many scientific applications
fit in and a very popular model for Grid Computing. Also, Task programming allows
the parallelization of legacy applications on the Cloud.
2.1.2 Concurrent Applications: Thread Programming Model
Thread Programming Model offers developers the capability of running multithreaded
applications on the Aneka Cloud. The main abstraction of this model is the concept of
thread which mimics the semantics of the common local thread but is executed
remotely in a distributed environment. This model offers finer control on the
execution of the individual components (threads) of an application but requires more
management when compared to Task Programming, which is based on a “submit and
forget” pattern. The Aneka Thread supports almost all of the operations available for
traditional local threads. More specifically an Aneka thread has been designed to
mirror the interface of the System.Threading.Thread .NET class, so that developers
can easily move existing multi-threaded applications to the Aneka platform with
minimal changes. Ideally, applications can be transparently ported to Aneka just by
substituting local threads with Aneka Threads and introducing minimal changes to the
code. This model covers all the application scenarios of the Task Programming and
solves the additional challenges of providing a distributed runtime environment for
local multi-threaded applications.
2.1.3 Data Intensive Applications: MapReduce Programing Model
MapReduce Programming Model [11] is an implementation of the MapReduce model
proposed by Google [12], in .NET on the Aneka platform. MapReduce has been
designed to process huge quantities of data by using simple operations that extracts
useful information from a dataset (the map function) and aggregates this information
together (the reduce function) to produce the final results. Developers provide the
logic for these two operations and the dataset, and Aneka will do the rest, making the
results accessible when the application is completed.
2.2 Overview of Windows Azure Platform
Generally speaking, Windows Azure Platform is a Cloud platform which provides a
wide range of Internet Services [3]. Currently, it involves four components (Figure 3).
They are Windows Azure, SQL Azure, Windows Azure AppFabric, and Windows
Azure Market Place respectively.
Windows Azure, which we will introduce in detail in Section 2.3, is a Windows based
Cloud services operating system providing users with on-demand compute service for
running applications, and storage services for storing data in Microsoft data centres.
The second component, SQL Azure offers a SQL Server environment in the Cloud,
whose features includes supporting Transact-SQL and support for the synchronization
of relational data across SQL Azure and on-premises SQL Server.
6
Figure 3: The Components of Windows Azure Platform.
Windows Azure AppFabric is a Cloud-based infrastructure for connecting Cloud and
on-premise applications, which are accessed through HTTP REST API.
The newly born Windows Azure Marketplace is an online service for making
transactions on Cloud-based data and Windows Azure Applications.
2.3 Overview of Windows Azure Service Architecture
In contrast to other public Cloud platforms such as Amazon EC2 and GoGrid,
Windows Azure currently does not provide an IaaS (Infrastructure-as-a-Service).
Instead, it provides a PaaS (Platform as a Service) solution, restricting users from
direct access with administrative privileges to underlying virtual infrastructure. Users
can only use the Web APIs exposed by Windows Azure to configure and use
Windows Azure services.
A role on Windows Azure refers to a discrete scalable component built with managed
code. Windows Azure currently supports three kinds of roles [4], as shown in Figure
4.
Web Role: a Web role is a role that is customized for Web application
programming as is supported by IIS 7.
Worker Role: a worker role is a role that is useful for generalized development.
It is designed to run a variety of Windows-based code.
VM Role: a virtual machine role is a role that runs a user-provided Windows
Server 2008 R2 image.
A Windows Azure service must include at least one role of either type, but may
consist of any number of Web roles, worker roles and VM roles. Furthermore, we can
launch any number of instances of a particular role. Each instance will be run in an
independent VM and share the same binary code and configuration file of the role.
Windows Azure
Windows Azure AppFabric
Applications and Data
Windows AzureMarketplace
7
Figure 4: Windows Azure Service Architecture.
In terms of the communication support, there are two types of endpoints that can be
defined: input and internal. Input endpoints are those are exposed to the Internet, and
internal endpoints are used for communication inside the application within the Azure
environment. A Web role can define a single HTTP endpoint and a single HTTPS
endpoint for external users, whilst a Worker Role and a VM role may assign up to five
internal or external endpoints using HTTP, HTTPS or TCP. There exists a built-in
load balancer on top of each external endpoint which is used to spread incoming
requests across the instances of the given role. Besides, all the role instances can make
outbound connections to Internet resources via HTTP, HTTPS or TCP.
Under this circumstance, we can deploy Aneka Container as instances of Windows
Azure Worker Role which gets access to resources on the Windows Azure
environment via the Windows Azure Managed Library.
2.4 Advantages of Integration of two platforms
Inevitably, the integrated Cloud environment will combine features from the two
platforms together, enabling the users to leverage the best of both platforms such as
access to cheap resources, easy programming, and management of Cloud computing
services.
2.4.1 Features from Windows Azure
For the users of Aneka Platform, the integration of the Aneka PaaS and Windows
Azure resources means they do not have to build or provision the infrastructure
needed for Aneka Cloud. They can launch any number of instances on Windows
Azure Cloud Platform to run their application in parallel to gain more efficiency.
Web Role InstancesWorker Role Instances
VM Role Instances
Load Balancer
User Applications
8
2.4.2 Features from Aneka Cloud Application Development Platform
For the users of Windows Azure Application, the integration of Aneka PaaS and
Windows Azure Platform allows them to embrace the advanced features from Aneka
PaaS:
Multiple Programming Models. As discussed in Section 2.1, the Aneka PaaS
provides users with three different kinds of cloud programming models, which
involves Task Programming, Thread Programming, and MapReduce
Programming to cover different application scenarios, dramatically decreasing
the time needed in developing Cloud-aware applications, as shown in Figure 5.
Figure 5: Multiple Programming Models of the Aneka PaaS Patent.
Scheduling and Management Services. The Aneka PaaS Scheduling Service
can dispatch the collection of jobs that compose an Aneka Application to the
compute nodes in a completely transparent manner. The users do not need to
take care of the scheduling and the management of the application execution.
Execution Services. The Aneka PaaS Execution Services can perform the
execution of distributed application and collect the results on the Aneka Worker
Container runtime environment.
Accounting and Pricing Services. Accounting and Pricing services of Aneka
PaaS enable billing the final customer for using the Cloud by keeping track of
the application running and providing flexible pricing strategies that are of
benefit to both the final users of the application and the service providers.
Dynamic Provisioning Services. In current pricing model for Windows Azure,
customers will be charged at an hourly rate depending on the size of the
compute instance. Thus it makes sense to dynamically add instances to a
deployment at runtime according to the load and requirement of the application.
Similarly instances can be dynamically decreased or the entire deployment can
be deleted when not being actively used to avoid charges. One of the key
features of Aneka is its support for dynamic provisioning which can be used to
leverage resources dynamically for scaling up and down Aneka Clouds,
controlling the lifetime of virtual nodes.
2.5 Limitations for the Integration
Although the integration of two platforms will generate numerous benefits for both
Aneka users and Windows Azure customers, running Aneka Container as instances of
Windows Azure Worker Role has some limitations.
The current version of Windows Azure does not provide administrative privileges on
Windows Azure Worker Role instances. Deployments are prepared using the
Aneka Container Core Services
Programming Model A: Task Programming
Aneka Container Core Services
Programming Model B: Thread Programming
Aneka Container Core Services
Programming Model C: MapReduce Programming
9
Windows Azure Managed Library, and the prepared Windows Azure Service Package
is uploaded and run.
Under these circumstances, we cannot use the Aneka Management Studio to install
Aneka Daemons and Aneka Containers on Windows Azure VMs directly. Further,
other third party software that is needed on the Worker nodes such as PovRay and
Maya, cannot be run on Windows Azure Worker Role instances because of the need
for administrative privileges. This limits the task execution services that Azure Aneka
Worker offers to XCopy deployment applications.
2.6 Challenges for the Integration
Due to the access limitations and service architecture of Windows Azure, we
encountered some implementation issues that required changes to some parts of the
design and implementation of the Aneka framework.
2.6.1 Administration Privileges
The Azure applications in both Web role and worker role instances do not have
administrative privileges and does not have write access to files under the
“E:\approot\” where the application code is deployed. On possible solution is to
use LocalResource to define and use the local resource of Windows Azure VM disk.
Technically speaking, we need to dynamically change the path of files which are to be
written to the local file system, to the path under the RootPath Property returned by
the LocalResource object at runtime.
2.6.2 Routing in Windows Azure
Each Windows Azure Worker Role can define up to five input endpoints using HTTP,
HTTPS or TCP, each of which is used as external endpoints to listen on a unique port.
One of the several benefits of using Windows Azure is that all the requests connected
to an input endpoint of a Windows Azure Role will be connected to a load balancer
which automatically forwards the requests to all instances that are declared to be part
of the role, on a round robin basis.
Figure 6: Routing in Windows Azure.
User Applications
Load Balancer Instance 1 of XX Role
Instance 2 of XX Role
Instance 3 of XX Role
1
2
First Request1
Subsequent Request2
10
As depicted in Figure 6, instances from the same Role will share the same defined
input endpoints and be behind the same Load Balancer. It is the responsibility of the
Load Balancer to dispatch the incoming external traffic to instances behind it
following a round robin mechanism. For instance, the first request will be sent to the
first instance of the given worker role, the second will be sent to the second available
instance, and so forth.
As we plan to deploy Aneka Container as instances of the Windows Azure Worker
Role, there exists a situation where the Aneka Master is outside the Windows Azure
Cloud and tries to send messages to a specific Aneka Worker inside the Windows
Azure Cloud. Since the load balancer is responsible for forwarding these messages,
there is a good possibility that the message may be sent to a Aneka Worker Container
other than the specified one. Hence, in order to avoid the message being transferred to
the wrong Aneka Worker Container, two possible solutions are available:
Forward Messages among Aneka Worker Containers. When a Container
receives a message that does not belong to it, it will forward the message to the
right Container according to the InternalEndpoint address encoded in the
NodeURI of Target Node of the Message. The advantage of this solution is the
consistency of the architecture of Aneka PaaS since no new components are
introduced to the architecture. The disadvantage, however, is that the
performance of Aneka Worker Containers will be hindered due to the overhead
of forwarding message.
Deploy a message proxy between the Aneka Worker Containers and
Master for the purpose of dispatching incoming external messages. The
Message Proxy is a special kind of Aneka Worker Container which does not
host any execution services. When the Windows Azure Aneka Cloud starts up,
all Aneka Worker Containers in Windows Azure encode the internal endpoint
address into the NodeURI. When the Message Proxy receives a message from
the Master, it dispatches the message to the right Aneka Worker Container
according to the encoded NodeURI specified in the Target Node of Message.
The disadvantage of this solution is that it costs extra since Windows Azure
charges according to the number of instances launched. However, in view of
possible performance issues, the second solution is preferred. More details on
the deployment of the Message Proxy Role are introduced in Section 3.1.3.
2.6.3 Dynamic Environment
As mentioned in Section 2.6.2, each Windows Azure Web Worker Role can define an
external endpoint to listen on a unique port. As a matter of fact, the port of endpoints
defined is the port of the load balancer. The Windows Azure Fabric will dynamically
assign a random port number to each instance of the given role to listen on.
Consequently, before starting the Container, we need to get the dynamically assigned
endpoint via RoleEnvironment.CurrentRoleInstance.InstanceEndpoints Property
defined in the Windows Azure Managed Library and save it to the Aneka
Configuration File so that the Container can bind the TCP channel to the right port.
Another change required by the dynamic environment of Windows Azure is that we
need to set the NodeURI of an Aneka Worker Container to the URL of the Message
Proxy and encode the internal endpoint of the Container into the URL. When the
Aneka Master sends a message to the NodeURI of an Aneka Worker Container, the
11
Message Proxy receives the message and forwards it to the right Aneka Worker
Container according to the internal endpoint address encoded in the NodeURI.
Furthermore, due to the dynamic nature of Windows Azure, we also need to guarantee
that the Load Balancer sends the message to the instance of Message Proxy Role only
if the message channel of the instance is ready and all the instances of Aneka Worker
Role start to send Heartbeat Message to the Aneka Master located on-premises, after
the deployment of the Message Proxy Role is finished.
2.6.4 Debugging
Debugging a Windows Azure Application is a bit different from debugging other
Windows .Net applications.
In general, after we install Windows Azure Tools for Visual Studio we can debug a
Windows Azure application locally when it is running in the Development Fabric
during the development stage. However, after the application has been deployed on
Windows Azure public Cloud, we cannot remotely debug the deployed application
since we do not have direct access and administrative privilege on Windows Azure
VMs.Fortunately, in June 2010 Windows Azure Tools + SDK, Windows Azure
provides us with a new feature that enables us to debug issues that occur in the Cloud
via IntelliTrace. With IntelliTrace debugging we can log extensive debugging
information for a role instance while it is running in Windows Azure. Subsequently,
we can use the IntelliTrace logs to step through the code from Visual Studio.
3. DESIGN
In this section, we will discuss the design decisions for deploying Aneka Containers
on Windows Azure as instances of Worker Role, how to integrate and leverage the
dynamic provisioning service of Aneka, and how to exploit the Windows Azure
Storage as a file storage system for Aneka PaaS in detail. The deployment includes
two different types. The first type is to deploy Aneka Worker Containers on Windows
Azure while the Aneka Master Container is run on local or on-premise resource. The
second type is to deploy the entire Aneka PaaS including Aneka Master Container and
Aneka Worker Containers on Windows Azure.
3.1 Deploying Aneka Workers on Windows Azure
3.1.1 Overview
Figure 7 provides an overall view of the deployment of Aneka Worker Containers as
instances of Windows Azure Worker Role.
As shown in the figure, there are two types of Windows Azure Worker Roles used.
These are the Aneka Worker Role and Message Proxy Role. In this case, we deploy
one instance of Message Proxy Role and at least one instance of Aneka Worker Role.
The maximum number of instances of the Aneka Worker Role that can be launched is
limited by the subscription offer of Windows Azure Service that a user selects. In the
first stage of the project, the Aneka Master Container will be deployed in the on-
premises private Cloud, while Aneka Worker Containers will be run as instances of
Windows Azure Worker Role. The instance of the Message Proxy Role is used to
transfer the messages sent from the Aneka Master to the given Aneka Worker.
In this deployment scenario, when a user submits an application to the Aneka Master,
the job units will be scheduled by the Aneka Master by leveraging on-premises Aneka
Workers, if they exist, and Aneka Worker instances on Windows Azure
12
simultaneously. When Aneka Workers finish the execution of Aneka work units, they
will send the results back to Aneka Master, and then Aneka Master will send the
result back to the user application.
Figure 7: The Deployment of Aneka Worker Containers as Windows Azure Worker Role
Instances.
3.1.2 Aneka Worker Deployment
Basically, we can deploy Aneka Containers of the same configuration as an Azure
Worker Role, since they share the same binary code and the same configuration file.
We can setup the number of instances of an Azure Worker Role to be launched in the
Windows Azure Service Configuration file, which represents the number of Aneka
Containers that will be deployed on the Windows Azure Cloud. And also we need to
setup the Aneka Master URI and the shared security key in the Windows Azure
Service Configuration file. When the instances of Aneka Worker Role are started up
by Windows Azure Role Hosting Process, we firstly update the configuration of the
Aneka Worker Container and start the Container program. After the container starts
successfully, it will connect to the Aneka Master directly.
3.1.3 Message Proxy
As for the issue we discussed in Section 2.6.2, in order to guarantee that messages are
transferred to the right target node specified by the Aneka Master, we need a
mechanism to route messages to a given instance. Therefore, we introduce a Message
Proxy between the Load Balancer and Aneka Worker instances. As shown in the
Figure 8, all the messages that are sent to the Aneka Worker Containers in Windows
Azure Public Cloud will be transferred to the external input endpoint of Message
Proxy Role. All the messages will be transferred to the load balancer of the input
endpoint①. The load balancer will transfer the messages to the instance of Message
Proxy Role②. In this case, we only launch one instance for Message Proxy Role. The
Message Proxy picks the incoming message, and parses the NodeURI of target node
to determine the internal address of the target node, and then forwards the messages to
Load Balancer
Aneka Worker Instance 2
Aneka Worker Instance 3
Aneka Worker Instance 1
Aneka Master
Aneka Worker Nodes(on premises)
Aneka / Azure Hybrid Cloud
User Applications
Internet
Private Cloud
13
the given Aneka Worker③. The Aneka Worker will handle the message and send a
reply message to Aneka Master directly④.
Figure 8: How the Message Proxy Works.
3.1.4 Dynamic Provisioning
Windows Azure provides us with programmatic access to most of the functionality
available through the Windows Azure Developer Portal via the Windows Azure
Service Management REST API. Using the Service Management API, we can manage
our storage accounts and hosted services.
Hence, by using the completely extensible Dynamic Resource Provisioning Object
Model of Aneka PaaS and Windows Azure Service Management REST API, we can
integrate Windows Azure Cloud resources into Aneka’s infrastructure and provide
support for dynamically provisioning and releasing Windows Azure resource on
demand.
Specifically, the Aneka APIs offer the IResourcePool interface and the
ResourcePoolBase class as extension points for integrating new resource pools. By
implementing the interface and extending the abstract base class, we can support
provisioning of Aneka Worker Containers on Windows Azure by following these
steps:
Use the CSPack Command-Line Tool to programmatically packet all the
binaries and the service definition file to be published to the Windows Azure
fabric, into a service package file;
Use the Windows Azure Storage Services REST API to upload the service
package to Windows Azure Blob;
Use the Windows Azure Service Management REST API to create, monitor and
update the status of the deployment of the Windows Azure Hosted Service.
Use the Windows Azure Service Management REST API to increase or
decrease the number of instances of Aneka Worker Containers to scale out or
scale in on demand;
Use the Windows Azure Service Management REST API to delete the whole
deployment of the Windows Azure Hosted Service when the provisioning
service is shutting down.
Load Balancer
Aneka Worker Instance 2
Aneka Worker Instance 3
Aneka Worker Instance 1
Aneka MasterAneka
Message Proxy
1 2
34
14
3.2 Deploying Aneka Cloud on Windows Azure
In the second deployment scenario, we deploy the Aneka Master Container as
instance of Windows Azure Worker Role. After finishing this step, we can run the
whole Aneka Cloud infrastructure on the Windows Azure Cloud Platform, as can be
seen from Figure 9.
3.2.1 Overview
In this scenario, users submit Aneka applications outside of the Windows Azure
Cloud and receive the result of the execution from Windows Azure Cloud. The
advantage of this structure is that it can dramatically decrease message transfer delay
since all the messages between the Aneka Master and Aneka Workers are transferred
within the same data centre of Windows Azure, and the cost of data transfer charged
by Windows Azure will reduce greatly as well.
Further, for data persistence requirements, the Aneka Master Container, can directly
use the Relational Data Service Provided by SQL Azure which would have higher
data transfer rates and of higher security since they are located in the same Microsoft
data centre.
3.2.2 Aneka Deployment in Azure
Figure 9 shows two types of roles being deployed on the Windows Azure Cloud: one
instance of Aneka Master Role hosting the Aneka Master Container, and at least one
instance of the Aneka Worker Role hosting the Aneka Worker Container. The Aneka
Master Container and Aneka Worker Containers interact with each other via an
internal endpoint, whilst the client and Aneka Master Container interact via an
external endpoint of the Aneka Master instance.
Figure 9: The Deployment of Aneka Master Container.
3.2.3 File Transfer System
In the current version of Aneka Cloud, the FTP protocol is used to transfer data files
from the client to the Aneka Master (or a separate Storage Container) and between the
Aneka Master and Aneka Worker Containers. However, due to the limitation of a
maximum of 5 networking ports allowed on each Windows Azure Role instance, we
Load Balancer
Aneka Worker Instance 2
Aneka Worker Instance 3
Aneka Worker Instance 1
Aneka MasterUser Applications
Windows Azure Storage
Application Data Files(Input / Output)
Application StatePersistence
Application FilesAccess
15
can no longer use the FTP service to support file transfers on the Windows Azure
Cloud. Instead, we can leverage Windows Azure Storage to support file transfers in
Aneka.
In general, as illustrated in Figure 10, two types of Windows Azure Storage will be
used to implement the Aneka File Transfer System: Blobs and Queues. Blobs will be
used for transferring data files, and Queues for the purpose of notification. When
Aneka users submit the application, if the transfer of input data files is needed, the
FileTransferManager component will upload the input data files to the Windows
Azure Blob and notify the start and end of the file transfer to Aneka’s Storage Service
via Windows Azure Queue. Similarly, the Aneka Worker will download the related
input data file from Windows Azure Blob, and the start and end of the file transfer
will be notified via Windows Azure Queue. When the execution of the work unit is
completed in the Aneka Worker, if the transfer of output data files is needed, the
FileTransferManager component of Aneka PaaS will upload the output data files to
the Windows Azure Blob to enable Aneka users to download from it.
Figure 10: Using Windows Azure Storage Blob and Queue for Implementation of Aneka
File Transfer System.
4. IMPLEMENTATION
In this section, we will explore the implementation details of the design we presented
in Section 3. Section 4.1 displays the class diagrams of the new and changed
components in Aneka PaaS. Next, Section 4.2 illustrates the configuration setting of
the deployments, whilst Section 4.3 demonstrates the designed life cycle of
deployments.
Aneka Master(Storage Service)
User Applications
Windows Azure Blob
Application Data Files(Input / Output)
Windows Azure Queue
File Transfer Notifications
Aneka Worker Instance
16
Figure 11: Class Diagram for Windows Azure Aneka Container Hosting Component.
Figure 12: Class Diagram for Windows Azure Aneka Provisioning Resource Pool
Component.
4.1 Class Diagrams
4.1.2 Windows Azure Aneka Container Deployment
Technically, in order to start an Aneka Container on Windows Azure Role instance,
we need to extend the RoleEntryPoint class which provides a callback to initialize,
run, and stop instances of the role. We override the Run() method to implement our
code to start the Aneka Container which will be called by Windows Azure Runtime
after the role instance has been initialized. Also worth noting is that due to the
dynamic nature of the Windows Azure environment, the configuration of Aneka
17
Worker Containers must be updated using the information obtained from the
CloudServiceConfiguration Class.
Figure 13: Class Diagram for Windows Azure Service Management Component.
4.1.2 Windows Azure Provisioning Resource Pool
The extendable and customizable Dynamic Resource Provisioning Object Model of
Aneka PaaS enables us to provide new solutions for dynamic provisioning in Aneka.
18
Specifically speaking, the WindowsAzureResourcePool class extends the
ResourcePoolBase class and implements the IResourcePool interface to integrate
Windows Azure as a new resource pool. The class WindowsAzureOperation provides
all the operations that are needed to interact with the Windows Azure Service
Management REST API.
4.1.3 Windows Azure Service Management
The DeploymentOperation component is used to interact with the Windows Azure
Service Management REST API to manage the Windows Azure Hosted Services in
terms of creating a deployment, updating the status of a deployment (such as from
Suspended to Running or vice versa) upgrading the deployment, querying the
deployment and deleting the deployment. This component is used by the Resource
Provisioning Service to manage the Windows Azure resource pool, and is also used
by the Windows Azure Role Deployment to monitor the status of deployment.
4.1.4 File Transfer System
The File Transfer System Component is used to transfer data files which are used in
application between clients and Aneka Cloud deployed on top of Windows Azure.
The class AzureFileChannelController which implements the IFileChannelController
interface represents the server component of the communication channel. It is
responsible for providing the connection string for the client component to gain access
to the Windows Azure Storage Service providing a way to upload and retrieve a
specific file. The class AzureFileHandler which implements the IFileHandler
interface is in charge of retrieving a single file or a collection of files from the server
component of the communication channel and uploading a single file or a collection
of files to the server component of the communication channel.
Figure 14: Class Diagram for Windows Azure Aneka Storage Service Implementation
using Windows Azure Storage.
19
4.2 Configuration
4.2.1 Provisioning Aneka Workers from Aneka Management Studio
In order to enable the Provisioning Service of Aneka to provision resources on
Windows Azure, we need to configure it via the Aneka Cloud Management Studio,
while configuring the services of the Aneka Master Container. This requires
configuring the Scheduling Service and Resource Provisioning Service.
Figure 15: Provisioning Service Configuration.
For the Scheduling Service, we need to select a proper scheduling algorithm for the
TaskScheduler and ThreadScheduler. Currently, only two algorithms are available for
dynamic provisioning: FixedQueueProvisioningAlgorithm and DeadlinePriority-
ProvisioningAlgorithm.
Figure 16: Windows Azure Resource Provisioning Service Configuration.
The configuration required for the Resource Provisioning Service in order to acquire
resources from the Windows Azure Cloud Service Providers is depicted in Figure 16.
For setting up a Windows Azure Resource Pool, we need the following information:
Capacity: identifies the number of instances that can be managed at a given
time by the pool. This value is restricted to the maximum number of instances
20
that the user is allowed to launch on Windows Azure, based on the
subscription.
Certificate File Path: specifies the file path of an X509 certificate that is used
to interact with Windows Azure Service Management REST API. This
certificate must also be uploaded to the Windows Azure Development Portal.
Certificate Password: designates the password of the X509 Certificate.
Certificate Thumbprint: assigns the thumbprint of the X509 Certificate.
Hosted Service Name: identifies the name of the Windows Azure Hosted
Service; the service must have been created via the Windows Azure
Development Portal.
Subscription ID: specifies the Subscription ID of Windows Azure Account.
Storage Account Name: designates the name of Windows Azure Storage
account that is under the same subscription.
Storage Account Key: specifies the key of the storage account.
Storage Container: defines the name of storage container which is used to
store the Windows Azure Hosted Service Package File.
Figure 17: Windows Azure Service Configuration File related to Windows Azure Aneka
Cloud Package.
4.2.2 Deploying Aneka Cloud on Windows Azure
In order to deploy an Aneka Cloud on Windows Azure, before uploading the
Windows Azure Aneka Cloud Package into Windows Azure Cloud, we need to
configure the Windows Azure Service Configuration file related to the Windows
Azure Aneka Cloud Package. To be more specific, as shown in the Figure 17, we
need to specify the values below:
DiagnosticsConnectionString: The connection string for connecting to the
Windows Azure Storage Service which is used to store diagnostics data.
21
DataConnectionString: The connection string for connecting to Windows
Azure Storage Service which is used to implement the File Transfer System.
SharedKey: The security key shared between Aneka Master and Aneka
Worker.
SubscriptionID: The Subscription ID of Windows Azure Account.
HostedServiceName: The name of the Windows Azure Hosted Service.
CertificateThumbprint: The thumbprint of the X509 Certificate which has
been uploaded to Windows Azure Service Portal. The value of thumbprint in
the Certificate Property should also be set.
AdoConnectionString: The connection string used to connect to an ADO
relational database if relational database is used to store persistent data.
More importantly, we need to define the Instance Number of Aneka Workers
running on Windows Azure Cloud, which is specified in the “count” attribute of
“Instance” property.
4.3 Life Cycle of Deployment
4.3.1 Aneka Worker Deployment
Figure 18 shows the whole life cycle of deployment of Aneka Worker Containers on
Windows Azure Cloud.
Configure Windows Azure
Resource Provisioning Pool
on Aneka Management Studio
Start Aneka Master Container
Start Provisioning Resource on
Windows Azure according to
the requirement of Application
Configuration for Installing
Aneka Master Container
on Aneka Management Studio
Configure
Scheduling Service Configuration
on Aneka Management Studio
Release Windows Azure
Resource when it is not actively
used within a given time block
Terminate the deployment of
Windows Azure Hosted
Service
Create the deployment of
Windows Azure Hosted Service
Update the status of
deployment From Suspended
to Runnint
Check and Delete current
deployment if it have
Add Windows Azure
Resource according to
requirement of Application
Configuration ReceiveFirst Request
for Provisioning
Receive Requests for
Provisioning
Receive Requests for
Releasing Resource
Provisioning Service is
shutting down
End
Start
Figure 18: The Life Cycle of Aneka Worker Container Deployment on Windows Azure.
Generally speaking, the whole life cycle of Aneka Worker Container deployment on
Windows Azure involves five steps. They are Configuration, First Time Resource
22
Provisioning, Subsequent Resource Provisioning, Resource Release, and Termination
of Deployment respectively.
Submit the application into
Aneka Cloud for execution
Configure the Windows Azure
Service Configuration File
Deployment Windows Azure
Aneka Cloud Package in Windows
Azure Service Portal
Stop and Delete the Windows
Azure Deployment
Configuration
& Deployment
Windows Azure
Deployment is
Completed
Application Execution
is Finished
End
Start
Figure 19: The Life Cycle of Aneka Cloud Deployment on Windows Azure.
4.3.2 Aneka Cloud Deployment
Figure 19 shows the life cycle for deploying an entire Aneka Cloud on top of
Windows Azure.
In general, the whole life cycle of Aneka Cloud deployment on Windows Azure
involves 3 steps. They are Configuration and Deployment, Application Execution, and
Deployment Termination respectively.
5. EXPERIMENTS
In this section, we will present the experimental results for application execution on
the Aneka Windows Azure Deployments including Aneka Worker Deployment and
Aneka Cloud Deployment. The test application we selected is Mandelbrot application
(Figure 20) which is developed on top of Aneka Thread Model to determine the
suitability of Aneka Windows Azure Deployment for running parallel algorithms.
Figure 21 displays the experimental results for executing the Mandelbrot application
using different input problem sizes, running on both Aneka Worker Deployment and
Aneka Cloud Deployment on Windows Azure when the number of Aneka Workers
being launched is 1, 5, and 10. The compute instance size of the Azure Instance
selected to run the Aneka Worker Containers is small computer instance which is a
virtual server with dedicated resources (CPU 1.6 GHz and Memory 1.75 GB) and
specially tailored for Windows Server 2008 Enterprise operating system as the guest
OS. The instance size for deploying the Aneka Master Container is medium computer
instance with machine configuration CPU 2*1.6 GHz and Memory 3.5 GB.
23
Figure 20: Mandelbrot Application developing on top of Aneka Thread Model.
Figure 21: Experimental Result Showing the Execution Time for Running Mandelbrot
Application on Aneka Worker Deployment.
From both Figure 21 and Figure 22, we can see that for the same input problem size,
there is a decrease in the execution time as a result of employing more Aneka
Workers to process the work units. The elapsed time used to execute application on
Aneka Worker Deployment is also much larger than on Aneka Cloud Deployment due
to the communication overhead between the Aneka Master and Aneka Workers with
Aneka Workers deployed inside Windows Azure Cloud, while Aneka Master is
deployed outside.
0
5
10
15
20
25
30
35
25 100 625 2500
1
5
10
Ap
plicatio
n Execu
tion
Time
(min
ute
s)
input problem sizes
num of workers
Experimental Result for Aneka Worker Deployment
24
Figure 22: Experimental Result Showing the Execution Time for Running Mandelbrot
Application on Aneka Cloud Deployment.
Figure 23: Scalability Diagram for Aneka Cloud Deployment.
In the next experiment, we measure the scalability of Aneka Cloud Deployment. In
this experiment, we use up to 16 small size instances. All the instances are allocated
statically. The result of the experiment is summarized in Figure 23. We see that the
throughput of the Mandelbrot application running on Azure Cloud Deployment
increases when the number of instances ascend.
0
5
10
15
20
25
25 100 625 2500
1
5
10
Ap
plicatio
n Execu
tion
Time
(min
ute
s)
input problem sizes
num of workers
Experimental Result for Aneka Cloud Deployment
84.80
126.60
201.96
277.90
424.86
1.00 1.49
2.38
3.28
5.01
0.00
1.00
2.00
3.00
4.00
5.00
6.00
0.00
50.00
100.00
150.00
200.00
250.00
300.00
350.00
400.00
450.00
1 2 4 8 16
# Aneka Worker Containers
Aggre
gated
Thro
ugh
pu
t (A
nekaTh
read
/min
)
Scalability
25
Figure 24: the Job Distribution Chart shown on the Aneka Analytics Tool
Furthermore, we can see from Figure 24 that the jobs are evenly distributed across all
the available Aneka Workers whose number is 10 in this case.
6. RELATED WORK
Windows Azure has been adopted in many projects to build high performance
computing applications [7, 8, 9]. Augustyn and Warchał [7] presented an idea and
implementation on how to use Windows Azure computing service to solve the N-
body problem using Barnes-Hut Algorithm. All computations are operated in parallel
on Windows Azure Worker Role instances. Lu et al. [8] delivered a case study of
developing AzureBlast, a parallel BLAST engine running on Windows Azure
Platform, which can be used to run the BLAST [13], a well-known and both data
intensive and computational intensive bioinformatics application. Li et al. [9]
demonstrated how to build the MODIS Satellite Data Reprojection and Reduction
Pipeline on Windows Azure.
In these cases, the whole implementation is started from scratch, which means the
developers need to handle application administration, task scheduling, communication
and interaction between role instances, and the storage service access. The Aneka
PaaS integration with Windows Azure Platform can speed up the entire development
for high performance application running on top of Windows Azure by using the
programming models powered by Aneka.
Besides, similar to Aneka, Lokad-Cloud [10] is an execution framework which
provides build-in features such task scheduling, queue processing and application
administration, and allows users to define a set of services to be run in Windows
Azure Worker Role instances. Nevertheless, different from the Aneka PaaS, Lokad-
Cloud is only designed to run applications on top of Windows Azure. It is worth
mentioning that Aneka PaaS is designed to run applications on private Cloud as well
as on public Clouds such as Windows Azure and Amazon EC2. Aneka PaaS can be
leveraged to integrate private Clouds with public Clouds by dynamically provisioning
resources on public Clouds such as Windows Azure when local resources cannot meet
26
the computing requirement. Moreover, Aneka supports three types of programming
models, the Task Model, Thread Model and MapReduce Model, to meet the
requirements of different application domains.
7. SAMPLE APPLICATIONS OF ANEKA
Different from other Cloud platforms or Cloud applications running on top of
Windows Azure we introduced in Section 6, Aneka allows seamless integration of
public Clouds and private Clouds to leverage their resources to executing
applications. Specifically, a wide range of applications from scientific applications,
business services, to entertainment and media, or manufacturing and engineering
applications have benefited from Aneka PaaS. A list application types that utilised
Aneka is shown in Table 1.
8. CONCLUSIONS AND FUTURE DIRECTIONS
In this chapter, we have introduced the Aneka Cloud Application Development
Platform (Aneka PaaS), presented and discussed the background, design and
implementation of the integration of the Aneka PaaS and Windows Azure Platform.
The Aneka PaaS is built on a solid .NET service oriented architecture allowing
seamless integration between public Clouds and mainstream applications. The core
capabilities of the framework are expressed through its extensible and flexible
architecture as well as its powerful application models featuring support for several
distributed and parallel programming paradigms. These features enhance the
development experience of software developers allowing them to rapidly prototype
elastically scalable applications. Applications ranging from the media and
entertainment industry, to engineering, education, health and life sciences and several
others have been proven to be appropriate to the Aneka PaaS.
Admittedly, the integration of two platforms would give numerous benefits to not
only the users of Aneka PaaS but also the customers of Windows Azure Platform,
enabling them to embrace the advantages of Cloud computing in terms of more
computing resources, easier programming model, and more efficiency on application
execution at lower expense and lower administration overhead.
In the first stage, we deployed the Aneka Worker Container as instances of Windows
Azure Worker Role, as well as support for dynamic provisioning of Aneka Workers
on Windows Azure. In the second step, we deployed the Aneka Master Container on
Windows Azure as an instance of Worker Role and the entire Aneka PaaS ran
completely on the Windows Azure Platform. This allows users to run Aneka Cloud
applications without requiring any local infrastructure. The message transfer overhead
and the transfer cost will decrease dramatically. This is beneficial to both Service
Providers who uses Aneka PaaS to deliver their services and the final users who
consume the services.
27
Table 1: Sample Applications of Aneka
Industry Sectors Challenges and Issues Aneka PaaS Usage
1. Geospatial
Sciences and
Technologies
More geospatial and non-spatial
data is involved due to increase
in the number of data sources
and advancement of data
collection methodologies.
Enable a new approach to complex
analyses of massive data and
computationally intensive
environments.
Build a high-performance and
distributed GIS environment over the
public, private and hybrid Clouds.
2. Health and Life
Sciences
High volume and density of data
require for processing. Enable faster execution and massive
data computation.
Suitable for life science R&D,
clinical simulation, and business
intelligence tools.
3. Financial
Services
Applications such as portfolio
and risk analysis, credit fraud
detection, option pricing require
the use of high-performance
computing systems and complex
algorithms.
Simplify the application
development lifecycle.
Reduce hardware investment.
Lower ongoing operational
expenditure.
Brings a breakthrough in industry
standard tools for financial
modelling such as Microsoft Office
Excel by solving its computational
performance barrier.
4. Telecom
Industry
The majority of Telecom
providers have several disparate
systems and they don’t have
enough capacity to handle the
utilization and access
information to optimize their
use.
Help telecom providers to realize
system utilization strategies in a cost
effective, reliable, scalable and
tightly integrated manner.
Help mission-critical applications by
automating their initiation across a
shared pool of computational
resources, by breaking the executions
into many parallel workloads that
produce results faster in accordance
with agreed upon SLAs and policies.
5. Manufacturing
and
Engineering
Manufacturing organizations are
faced with a number of
computing challenges as they
seek to optimize their IT
environments, including high
infrastructure costs and
complexity to poor visibility into
capacity and utilization.
Enable organizations to perform
process simulation, modelling, and
optimization at a highly increased
rate so that the time-to-market of key
products is faster, by effectively
leveraging Cloud technologies.
6. Entertainment
and Media
Business solutions involving
digital media transcoding to HD
video, 3D image rendering, and
gaming, require plenty of time to
process and utilize vast amounts
of computing capacity to encode
and decode the media.
Optimize networked computers as a
private Cloud or leverage public
Cloud such as Windows Azure,
Amazon EC2 and Go-Grid.
Allows scaling applications linearly.
Better utilize the Cloud farm
providing best efficiency and speed
possible using Cloud scalability.
28
On the whole, in addition to the integration with Windows Azure Platform, presently,
Aneka PaaS has already supported the integration of Amazon EC2, GoGrid, and Xen
Server. The support of provisioning resources on Windows Azure Platform once
again illustrates the adaptability, flexibility, mobility and extensibility of the Aneka
PaaS. In the next stage, the Aneka PaaS will continue to integrate with other public
Cloud platforms and virtual machine management platforms such as VMWare,
Microsoft HyperV and so forth, to help users to exploit more power of Cloud
computing.
ACKNOWLEDGEMENT
This work is partially supported by a grant from the Australian Department of
Innovation, Industry, Science and Research via its COMET (Commercialising
Emerging Technologies) Program.
REFERENCES
[1] Christian Vecchiola, Xingchen Chu, and Rajkumar Buyya, Aneka: A Software
Platform for .NET-based Cloud Computing, High Speed and Large Scale
Scientific Computing, 267-295pp, W. Gentzsch, L. Grandinetti, G. Joubert (Eds.),
ISBN: 978 -1-60750-073-5, IOS Press, Amsterdam, Netherlands, 2009.
[2] Rajkumar Buyya, Chee Shin Yeo, Srikumar Venugopal, James Broberg, and Ivona
Brandic, Cloud Computing and Emerging IT Platforms: Vision, Hype, and
Reality for Delivering Computing as the 5th Utility, Future Generation Computer
Systems, 25(6):599-616, Elsevier Science, Amsterdam, The Netherlands, June
2009.
[3] David Chappell, Introducing the Windows Azure Platform. David Chappell &
Associates, October 2010.
[4] David Chappell, Introducing Windows Azure. David Chappell & Associates,
October 2010.
[5] Henry Li, Introducing Windows Azure, ISBN: 978-1-4302-2469-3, Apress, 2009.
[6] Sriram Krishnan, Programming Windows Azure: Programming the Microsoft
Cloud, ISBN: 978-0-596-80197-7, O’REILLY, Sebastopol, CA, USA, May 2010.
[7] Dariusz Rafał Augustyn and Łukasz Warchał, Cloud Service Solving N-Body
Problem Based on Windows Azure Platform, Proceedings of the 17th
Conference
on Computer Networks, Communications in Computer and Information
ScienceUstron, Poland, June 15-19, 2010.
[8] Wei Lu, Jared Jackson, and Roger Barga, AzureBlast: A Case Study of Developing
Science Applications on the Cloud, Proceedings of the 19th ACM International
Symposium on High Performance Distributed Computing, Chicago, Illinois, USA,
June, 2010.
[9] Jie Li, Marty Humphrey, Deb Agarwal, Keith Jackson, Catharine van Ingen and
Youngryel Ryu, eScience in the Cloud: A MODIS Satellite Data Reprojection
and Reduction Pipeline in the Windows Azure Platform, Proceedings of the 24th
29
IEEE International Symposium on Parallel & Distributed Processing (IPDPS
2010), Atlanta, GA, USA, April, 2010.
[10] Lokad.Cloud Project, http://code.google.com/p/lokad-cloud/.
[11] Chao Jin, Rajkumar Buyya, MapReduce Programming Model for .NET-Based
Cloud Computing, Euro-Par 2009: 417-428.
[12] Sanjay Ghemawat and Jeffrey Dean, MapReduce: Simplified Data Processing on
Large Clusters, Symposium on Operating System Design and Implementation
(OSDI’04), San Francisco, CA, USA, 2004.
[13] S.F. Altschul, W. Gish, W. Miller, E.W. Myers, and D.J. Lipman, Basic Local
Alignment Search Tool, Journal Molecular Biology, 215(3): 403–410, 1990.
doi:10.1006/jmbi.1990.9999.
[14] Michael Armbrust, Armando Fox, Rean Griffith, Anthony D. Joseph, Randy
Katz, Andy Konwinski, Gunho Lee, David Patterson, Ariel Rabkin, Ion Stoica,
Matei Zaharia, A View of Cloud Computing, Communications of the ACM, Vol.
53 No. 4, Pages 50-58, 2010.